Planning sustainable electricity solutions for refugee settlements in sub-Saharan Africa

Building a comprehensive database

The RSEA DB was designed and developed to gather the multidimensional factors47 that determine the energy access dimension for refugee settlements. These factors were chiefly environmental (such as variability of solar radiation and avoided GHG emissions), technical (electrification status, distance to grid), social (population, electricity demand, social infrastructure) and economic (PV mini-grid component prices, discount factors, operation and maintenance costs). The main indicators capturing these factors are listed in Supplementary Table 6.

To retrace the steps behind the creation of the RSEA DB, note the following.

  1. (1)

    The holistic structure of the RSEA DB was informed by an extensive review of the existing literature including whitepaper reports1,10,11,12,13,14,21,32,38,48,49 and academic papers3,4,5,6,8,9,19,25,34,35,36,41,42,45,46,50,51 which focus on energy access and refugee settlements. The analysed documents were used to gather demographic figures from UNHCR statistics, to obtain field data from humanitarian contexts and other relevant data from studies on solar and hybrid (solar-diesel) mini-grids installed in refugee camps or rural communities in Africa. Studies on mini-grids in similar contexts were analysed in depth and technical and financial information, such as generation component prices, battery costs and discount rates were retrieved, as well as documentation on the institutions operating in the camps, number of businesses, health facilities and so on. Additional information regarding the hosting countries was also collected from literature (for example, inflation rates, currency exchange, electricity tariffs).

  2. (2)

    Field data on electricity needs were collected and retrieved from the Kalobeyei settlement in Kenya in 2020. Specifically, current electricity consumption was measured from the existing electricity supply: diesel generators, an existing PV mini-grid (metered data), solar home systems (PV and battery storage) and PV panels7 already owned by refugees (second-hand panels were mostly found to be in poor conditions). The data collection was conducted via a questionnaire administered to the refugee population of the settlement. The semistructured survey reached 325 interviewees, properly selected among type of end-users (households, businesses and institutions) and their geographic location inside the camp (the settlement is divided into three villages).

  3. (3)

    The data collection was complemented by stakeholder interviews. The surveyed experts were selected from various humanitarian organizations (Danish Refugee Council, Norwegian Refugee Council and UNHCR), development actors (Deutsche Gesellschaft fur Internationale Zusammenarbeit, Global Plan of Action, Netherlands Development Organisation and United Nations Institute for Training and Research), research institutions (École Polytechnique Fédérale de Lausanne and Imperial College London) and civil society organizations (Renewvia Energy Kenya Limited) involved in the coordination and organization of refugee settlements in SSA. A complete list of the stakeholders interviewed is stored in Supplementary Table 7 (the response rate was around 10%). Selection of the questions used for the interviews was mainly based on the background, expertise and agency-provenience of the interviewees. Qualitative discussions with the experts were aimed to cross-check and validate the information from the literature, understand the socio-economic framework in refugee settlements in Africa (for example, electrification conditions, social context and refugee energy routines), to obtain verifiable data sources inherent to energy-related humanitarian operations and to complete the technical data input for actual energy operations based on field data from the settlements (system losses, efficiency, lifetime of the system and so on).

After completing the data collection and the techno-economic analysis described below, the following parameters were integrated in the RSEA DB for each of the settlements.

  1. (1)

    Descriptive parameters: name and coordinates of the site location, name and type of the refugee settlement, number of refugee population, actual or estimated number of households, businesses and public institutions. Additional parameters are distance to the national electric grid and the closest national border.

  2. (2)

    Energy demand parameters: residential, business and institutional daily electricity demand at settlement level (kWh d–1), estimated at year 5 of the lifetime of the project (20 yr). Daily electricity demand estimated at year 0 for a single household, business and public institution (kWh per connection per day).

  3. (3)

    Techno-economic parameters: optimized PV and battery size. Up-front cost of generation (PV, battery and balance of system). Up-front cost of distribution (customer connection fee, overhead lines and substation). Soft cost for logistics and project management and for contingencies. Replacement cost (inverter and battery). LCOE values with three discount rates (5%, 10% and 12%). LCOE sensitivity analyses for operation and maintenance costs (2% and 3% of up-front costs) and four levels of grant coverage of the customer connection fee (25%, 50%, 75% and 100%).

  4. (4)

    Climate parameters: daily and yearly avoided GHG emissions at year 5 of the lifetime of the project.

  5. (5)

    Geospatial datasets: PVGIS15, GHSL17, NOOA52, OSM53 and national grid54 were integrated in the RSEA DB to geographically contextualize the refugee settlements (distribution of population, solar radiation, accessibility maps, existence of electricity grid and night-light imagery).

The RSEA DB tracks 288 settlements organized in 203 refugee sites. The settlement coincides, in most cases, with the site. However, for some sites hosting a large number of refugees, population data were publicly available for smaller administrative units (for example, Bidibidi in Uganda, is organized in five units or ‘zones’ and Kakuma, in Kenya, in four ‘villages’). Hence, each subadministrative unit within the same site was traced as a different settlement. Conversely, in the absence of information regarding the subadministrative units, some large refugee sites had to be considered as a single settlement (for example, Nyarugusu in Tanzania).

Estimating electricity demand and daily load profiles

The electricity demand and load profile per refugee settlement were calculated by considering the three types of users below.

  1. (1)

    Residential users: the electricity demand at the household level corresponds to a small-scale power of needs (for example, for tier 1, task light and phone charging; and, for tier 2, adding general lighting, television, fan or computing).

  2. (2)

    Productive/ business users: the electricity demand for income-generating activities includes energy requirements for commercial usage and industrial light (small businesses).

  3. (3)

    Public/ institutional users: this category consists of demand for facilities such as street lighting, back-office administration, communication services, healthcare and education services.

Note that only users located inside the settlement were considered. For instance, humanitarian compounds located outside the camps are not included. The number of household users in each settlement was retrieved from UNHCR statistics or estimated from the average number of people per household of each country. The number of businesses and institutions were retrieved from the literature or expert interviews (nine settlements only) or estimated using a fixed ratio of business and institutional users to residential ones (respectively, 0.0507 and 0.0059, computed as average values across the nine observed settlements)—in fact, the number of businesses and institutions will vary (perhaps significantly) across settlements.

Settlement-specific daily load profiles for households, businesses and institutions were derived from the corresponding daily load profiles estimated for Kalobeyei. The latter process involved assessing, during the field visit, the hourly electricity demand of the respondents having access to any source of electricity in each village and adjusting these figures to the actual number of users. The available sources of electricity were an operational mini-grid (7% of the respondents), several diesel generators (9%) and solar lamps or second-hand PV solar panels purchased from the market (45%)—the remaining 61% of the respondents had no access to electricity. The estimated daily load profiles per type of user and per village in Kalobeyei are shown in Supplementary Table 8b.

To calculate, for each user type, a settlement-specific daily load profile, five corrective factors (CFs) were defined and applied to the load profiles in Supplementary Table 8b. These factors adjust the load profiles estimated for Kalobeyei, to the characteristics of each settlement and account for: the number of connections per type of user (CF1); the presence of users opting to remain with a pre-existing electricity provider (CF2); the fact that part of the users in Kalobeyei were already connected to a mini-grid (CF3); and the different tiers (CF4 and CF5 for, respectively, tier 2 and tier 3). A full description of the CFs can be found in Supplementary Table 8a. The aggregated daily load profile for each settlement is the sum, over all user types, of the settlement-specific daily load profiles per type of user—for an example see Supplementary Fig. 1.

Note that for the purpose of sizing the mini-grid capacity, the hourly power demand which composes the settlement-specific daily load profile per user type, is assumed to grow annually by 10%, starting from the initial values estimated for year 0 with the CFs and until year 5. After year 5, the hourly power demand remains constant for the lifetime of the project. These choices mimic the assumptions on potential demand growth made by mini-grid developers in Kenya. Expert interviews confirmed that the limited available experience and the uncertainty regarding the number of refugees living in each settlement can strongly affect the reliability of long-term forecasting for these projects.

The aggregate settlement-specific electricity demand and the settlement-specific electricity demand per user type, were calculated from the load profiles. Hence, the daily electricity demand for a single household depends on the year and on the scenario:

  1. (1)

    Tier 1—this scenario was based on the data registered in Kalobeyei (0.07 kWh d–1 per household in year 0).

  2. (2)

    Tier 2—this scenario is coherent with the SDG 7 framework20,24 and with the goal set by UNHCR21,51 for refugee settlements (0.20 kWh d–1 per household in year 0).

  3. (3)

    Tier 3—this scenario captures a potential increase from the current goal (tier 2) to the equivalent of tier 3 (ref. 38) access (1 kWh d–1 per household in year 0).

The daily electricity demand for a single business and for a single institution depend only on the year and were also derived from the field data collected in Kalobeyei (0.26 kWh d–1 and 3.14 kWh d–1, respectively, in year 0).

Optimizing the PV and hybrid PV/diesel mini-grid

The total amount of solar irradiation and the intermittency of the solar irradiation depend strongly on the geographical location. On the basis of these data and the specific load characteristics of each refugee settlement, the PV mini-grid was sized with 100% renewable energy and with PV plus diesel generator as backup. The optimal size of the PV array and battery storage for a given settlement was computed to ensure loss of power supply probability37 at 10%.

The annual electricity generated yearly by the PV mini-grid was calculated for each settlement using a combination of models20: (1) a model for effective irradiance with hourly solar radiation measured from satellites (combined with temperature and wind speed data from re-analysis); (2) a PV output power model with measured data on module performance; and (3) a model for battery performance based on measured battery data. This methodology has been described in detail and validated19,20,55. As for the hybrid system56,57, current diesel cost were used, differentiated per country, thus they include national diesel subsidies.

The distribution network consists of low-voltage overhead feeders sustained by poles and end-user connections. Settlements with a number of households above 2,000 (60% of the observed ones) include a substation as an additional distribution element.

Assessing electrification costs

The cost of providing electricity to each settlement via a PV plus battery storage mini-grid was analysed in two ways: (1) estimating an up-front capacity cost on the basis of the system specifications; and (2) computing the corresponding LCOE. The up-front capacity cost was calculated on the basis of the PV plus battery storage mini-grid optimization outputs (PV size and battery size) and three main cost categories: generation, distribution and soft costs. Field data collected in Kenya were used to set these three cost categories7 (Supplementary Table 1).

For each refugee settlement (n), the LCOE was calculated using a 20 years lifetime (T) and a discount rate (r) of 10% (refs. 15,58). The up-front capacity costs (Up_cost0 in year t = 0), the annual operation and maintenance cost (O&Mt), the asset replacement (Rt, battery and PV inverter at year 10) and financial costs (Ft) were taken into consideration by the following equation:

$$\mathrmLCOE_n = \frac{{\mathrmUp\_\mathrmcost_0 + \mathop \sum\nolimits_t = 1^T {\left\{ {\left( {\mathrmO\& \mathrmM_t + R_t + F_t} \right)/\left( 1 + r_n \right)^t} \right\}} }}{{\mathop \sum\nolimits_{{t} = 1}^T {\left\{ {\left( {\mathrmES_n} \right)/\left( {1 + r_{{n}}} \right)^{{{t}}}} \right\}} }}$$


where O&Mt costs are 1% of the up-front capacity costs, financial expenditures include VAT payments, insurance, interest expenses and \(\mathrmES_n\) is the annual electricity production from the PV mini-grid. The same financing structure used in Kalobeyei was assumed, with 90% grant, 8% equity and 2% debt.

The indicator LCOEn includes the entire up-front capacity costs. The indicator LCOE_genn (also reported in the database) is the part of the LCOE for the generation components only. The reference values and variations of the parameters used in the sensitivity analyses (discount rate, O&M costs and customer connection fee) are shown in Supplementary Table 5a. Supplementary Fig. 2 illustrates (for the reference tier 2 demand scenario) that, starting from O&M costs equal to 1% of the up-front costs59, an increase to 2% and 3% corresponds to a LCOE increases of 1.5% and 1.7%, respectively. In a similar manner, assuming that the customer connection fee might be covered partially (25%) or entirely (100%) by a grant also lead to relatively small decrease in the median value of the LCOE (comprising between –0.1% and –1.3%).

The effects on the LCOE values (under tier 2) of varying the discount rate were illustrated in Fig. 4a. To study the effect of a country-specific discount rate on the LCOE (Fig. 4b), weighted average capital cost values were calculated47 with recent input data on equity rate of return and debt interest rate in each country (World Bank/ IMF lending interest rate, 2021)60 using the methodology described previously40. The country-specific weighted average capital cost values (Supplementary Table 5b) were then included in the LCOE location-specific estimates for both the fully renewable and the hybrid PV/ diesel mini-grids56,57.

Estimating avoided emissions

The estimation of the carbon mitigation potential of using fully renewable mini-grids is based on the avoided GHG emissions in CO2e (ref. 59). The avoided GHG emissions were calculated by computing the annual emissions of a stand-alone diesel generator supplying the settlements’ electricity demand in year 5, when demand becomes stable and the system is operating at full capacity. Supplementary Table 9 summarizes the relevant parameters and sources used to compute the emission factor (0.93 tCO2 MWh–1).

Observing the host communities

The lack of harmonized geodata concerning the African population, together with inconsistent demographic information produced through census campaigns, has prompted the scientific community to extract information from the available satellite remote sensing archives61. Consistently, a multicriteria site-selection algorithm was designed to identify the populated areas at a given motorized travelling distance from 13 selected refugee settlements and without access to electricity. The criteria used to delineate these areas are:

  1. (1)

    Identify the borders of the selected refugee settlements by combining cartographical (OSM) and earth observation data. Missing boundaries were added through visual interpretation of satellite images (Microsoft Bing aerial images).

  2. (2)

    Estimate the cumulative travel distance by motorized44 means for the selected refugee settlements. The calculation considers the travel distance starting from the settlement borders until reaching the maximum distance of 10 km. The areas inside each site were not considered in this analysis.

  3. (3)

    Extract the catchment area considering the population without access to electricity18 around the refugee settlements that corresponds to the total population living inside these settlements.

The requirement of data extraction for obtaining missing settlement boundaries limited the analysis to a few settlement areas. This limitation can be overcome in future analyses if cartographic information becomes available or a specific algorithm is developed to automate the process of settlement boundary delimitation and eliminate the need for visual interpretation techniques.